About Companion Labs
We raised $2.5Mn in seed funding led by Peak XV Surge. We're building chat-based RPGs for India. Think web series, but you're the hero and AI plays every character.
The Role
Our entire product is LLM-powered.
Every character, every conversation, every scene.
If the AI doesn't feel real, the product doesn't work.
You'd own the prompt architecture and LLM systems that make our stories feel alive.
You're not building AI as a feature. AI is the product.
What you'll do
Design and own the prompt architecture: multi-turn dialogue, scene transitions, character voice, memory management across long sessions
Build agentic systems that orchestrate story flow. Deciding what happens next, when to escalate drama, when to introduce new characters.
Own model selection and routing. Pick the right model for the right task, optimise for cost, latency, and output quality.
Build eval pipelines. Measure character consistency, dialogue quality, narrative coherence. Catch regressions before users do.
Set up observability. Logging, tracing, debugging LLM behaviour in production.
Work directly with founders on product and technical architecture.
You're a fit if
You've worked with LLMs in production. Not just API calls, but understanding model behaviour, failure modes, and how to control outputs
You've built some form of agents. Multi-step, tool-using, autonomous or semi-autonomous systems.
You think in prompt chains, orchestration layers, and fallback logic. Not one-off prompts.
You're comfortable with Python, and have worked with frameworks like LangChain, LangGraph, or similar.
You've built with Claude Code, Cursor, or similar AI-native dev tools.
You care about what the user experiences, not just what the system outputs.